Have the SciML linter allow non-yaml network formats

petab/v2/extensions/sciml.py

@@ -272,21 +272,22 @@ def from_config(
         sciml_config: SciMLConfig = config.extensions[C.EXT_ID_SCIML]
 
         # Neural network classes are constructed via pytorch for now to get
-        # the proper inputs
-        neural_networks = [
-            NNModel.from_pytorch_module(
-                NNModelStandard.load_data(
-                    _generate_path(
-                        file_path=nn_config.location,
-                        base_path=base_path,
+        # the proper inputs. Non-YAML formats are opaque — the file is assumed
+        # to contain a valid model and is not read here.
+        neural_networks = []
+        for nn_id, nn_config in (sciml_config.neural_networks or {}).items():
+            if nn_config.format.lower() == "yaml":
+                neural_networks.append(
+                    NNModel.from_pytorch_module(
+                        NNModelStandard.load_data(
+                            _generate_path(
+                                file_path=nn_config.location,
+                                base_path=base_path,
+                            )
+                        ).to_pytorch_module(),
+                        nn_model_id=nn_id,
                     )
-                ).to_pytorch_module(),
-                nn_model_id=nn_id,
-            )
-            for nn_id, nn_config in (
-                sciml_config.neural_networks or {}
-            ).items()
-        ]
+                )
 
         hybridization_tables = [
             HybridizationTable.from_tsv(f, base_path)

petab/v2/extensions/sciml_lint.py

@@ -16,9 +16,11 @@ def run(self, problem: core.Problem) -> lint.ValidationIssue | None:
         condition_targets = {
             c.target_id for ct in problem.conditions for c in ct.changes
         }
+        # Only YAML-format networks are loaded as NNModel objects
         nn_input_ids = {
             inp.input_id
             for nn in problem.extensions.sciml.neural_networks
+            if hasattr(nn, "inputs")
             for inp in nn.inputs
         }
         hyb_target_ids = {

tests/v2/test_sciml.py

@@ -138,3 +138,16 @@ def _get_test_problem():
 def test_lint():
     problem = _get_test_problem()
     assert problem.validate() == []
+
+
+def test_lint_equinox_network_format():
+    """Linter accepts non-YAML formats without reading the network file."""
+    problem = _get_test_problem()
+    # Replace the YAML network config with equinox format
+    sciml_cfg = problem.config.extensions["sciml"]
+    sciml_cfg.neural_networks["net1"] = NeuralNetConfig(
+        location="net1.py",
+        pre_initialization=False,
+        format="equinox",
+    )
+    assert problem.validate() == []